Selection of representative protein data sets

Abstract

The Protein Data Bank currently contains about 600 data sets of three-dimensional protein coordinates determined by X-ray crystallography or NMR. There is considerable redundancy in the data base, as many protein pairs are identical or very similar in sequence. However, statistical analyses of protein sequence-structure relations require nonredundant data. We have developed two algorithms to extract from the data base representative sets of protein chains with maximum coverage and minimum redundancy. The first algorithm focuses on optimizing a particular property of the selected proteins and works by successive selection of proteins from an ordered list and exclusion of all neighbors of each selected protein. The other algorithm aims at maximizing the size of the selected set and works by successive thinning out of clusters of similar proteins. Both algorithms are generally applicable to other data bases in which criteria of similarity can be defined and relate to problems in graph theory. The largest nonredundant set extracted from the current release of the Protein Data Bank has 155 protein chains. In this set, no two proteins have sequence similarity higher than a certain cutoff (30% identical residues for aligned subsequences longer than 80 residues), yet all structurally unique protein families are represented. Periodically updated lists of representative data sets are available by electronic mail from the file server "netserv@embl-heideIberg.de." The selection may be useful in statistical approaches to protein folding as well as in the analysis and documentation of the known spectrum of three-dimensional protein structures.

One- to Four-Dimensional Kernels for Virtual Screening and the Prediction of Physical, Chemical, and Biological Properties

“…Redundancy reduction in prediction is common practice in areas such as bioinformatics but has not been applied systematically to data sets in chemoinformatics. To study and avoid the biases introduced by redundant data, we use the algorithm in Hobohm et al 35 to derive redundancy-reduced data sets, by iteratively thinning clusters of molecules with high similarity, until no molecules in the training sets have a similarity greater than some preset threshold (see Results, section 4.7).…”

Enlarged representative set of protein structures

“…The new representative list of PDB chain identifiers was produced using "algorithm 2" of Hobohm et al (1992). This algorithm removes redundant protein chains 1 by 1, following a strategy called "greedy" by computer scientists: the chain with the largest number of neighbors is removed, until no neighbors are left.…”

“…Although the wealth of data bears witness to the progress achieved by protein crystallographers and NMR spectroscopists, an overview of the spectrum of known protein structures and certain statistical analyses of protein structures require nonredundant data. To meet this need, we have developed algorithms to select from PDB (or from sequence databases) representative subsets that aim to minimize redundancy and maximize coverage (Hobohm et al, 1992). The result, in the form of a list of PDB identifiers, was published about 1; years ago and has been very useful, e.g., in developing better algorithms for secondary structure prediction (Rost & Sander, 1993).…”

“…The current "25%-list" (based on PDB release, December 1993), i.e., a list of protein Reprint requests to: Uwe Hobohm, European Molecular Biology Laboratory, 69012 Heidelberg, Germany; e-mail: hobohm@embl-heidelberg.de. chains with less than 25% sequence identity, was constrained to be downward compatible with the "30% list" published in 1992 (Hobohm et al, 1992). Some adjustments were a result of the more stringent threshold and new quality criteria, e.g., replacement of a data set by one with identical sequence but higher resolution.…”

Structural analysis based on state‐space modeling

“…Abola et al, 1987), comprising over 100 nonhomologous structures (Hobohm et al, 1992), it is well known that a number of structural themes recur. One example of such a recurring structure is the Rossman fold (Rossman et al, 1974), found in such proteins as lactate dehydrogenase, glyceraldehyde-3-phosphate dehydrogenase, and alcohol dehydrogenase (Branden & Tooze, 1991).…”